Comment by altairprime

Perhaps closer to “the mean vector point such that all outbound vectors to different training tests are in sum the smallest”? I assume that’s a property of neural networks anyways, though I’m out of date on current math for them.